README.md

metadata

license: apache-2.0
tags:
  - moe
  - frankenmoe
  - merge
  - mergekit
  - lazymergekit
  - LLM360/AmberChat
  - wannaphong/han-llm-7b-v2
base_model:
  - LLM360/AmberChat
  - wannaphong/han-llm-7b-v2

HanAmber-7b-MOE

HanAmber-7b-MOE is a Mixure of Experts (MoE) made with the following models using LazyMergekit:

• LLM360/AmberChat
• wannaphong/han-llm-7b-v2

🧩 Configuration

base_model: LLM360/AmberChat
dtype: float16
gate_mode: cheap_embed
experts:
  - source_model: LLM360/AmberChat
    positive_prompts: ["You are an helpful as general-pupose assistant."]
  - source_model: wannaphong/han-llm-7b-v2
    positive_prompts:
      - "คุณช่วยฉันหน่อยได้ไหม"
      - "คุณช่วยแปลประโยคนี้เป็นภาษาไทยได้ไหม"

💻 Usage

!pip install -qU transformers bitsandbytes accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "Manichik/HanAmber-7b-MOE"

tokenizer = AutoTokenizer.from_pretrained(model)
pipeline = transformers.pipeline(
    "text-generation",
    model=model,
    model_kwargs={"torch_dtype": torch.float16, "load_in_4bit": True},
)

messages = [{"role": "user", "content": "Explain what a Mixture of Experts is in less than 100 words."}]
prompt = pipeline.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
print(outputs[0]["generated_text"])